Video Generation on VBench 2.0
0.951Human FidelityPhyMotion
Evaluation Results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| PhyMotionTraining Strategy=Structured 3D reward training, Base Model=Causal Forcing 1.3B2026.05 | 0.951 | — | — | — | 0.411 | — | 0.9956 | — | 0.5589 | 0.9957 | |
| Wan2.2 14BModel Scale=14B2026.05 | 0.943 | — | — | — | 0.393 | — | 0.9892 | — | 0.5141 | 0.9857 | |
| VideoAlign-MQTraining Strategy=VLM-reward trained2026.05 | 0.91 | — | — | — | 0.57 | — | 0.9904 | — | 0.5229 | 0.9835 | |
| Causal Forcing 1.3BModel Scale=1.3B2026.05 | 0.886 | — | — | — | 0.631 | — | 0.9895 | — | 0.5391 | 0.9819 | |
| FastWan 1.3BModel Scale=1.3B2026.05 | 0.876 | — | — | — | 0.928 | — | 0.9847 | — | 0.5035 | 0.97 | |
| PhyMotionTraining Strategy=Structured 3D reward training, Base Model=FastWan 1.3B2026.05 | 0.867 | — | — | — | 0.42 | — | 0.9927 | — | 0.5218 | 0.9928 | |
| VPOOptimization Strategy=Prompt Optimizer, Base Model=CogVideoX-5B2026.03 | 0.8445 | 0.4444 | 0.5419 | 0.2479 | 0.3901 | 43.34 | — | — | — | — | |
| EchoMotion 5BModel Scale=5B2026.05 | 0.836 | — | — | — | 0.683 | — | 0.985 | — | 0.4756 | 0.9777 | |
| Wan 1.3BModel Scale=1.3B2026.05 | 0.834 | — | — | — | 0.7 | — | 0.9868 | — | 0.4758 | 0.9787 | |
| Wan2.2 5BModel Scale=5B2026.05 | 0.823 | — | — | — | 0.598 | — | 0.9846 | — | 0.4622 | 0.9803 | |
| Veo3 FastType=Proprietary system2026.02 | 0.822 | 0.556 | 0.619 | 0.415 | 0.679 | — | — | — | — | — | |
| Gemini-3-Pro (VQQA)Optimization Strategy=VQQA, Optimization rounds=4, Base Model=CogVideoX-5B2026.03 | 0.8123 | 0.5485 | 0.5821 | 0.315 | 0.5426 | 50.41 | — | — | — | — | |
| Wan 2.2-5BScale=5B2026.02 | 0.803 | 0.503 | 0.581 | 0.324 | 0.662 | — | — | — | — | — | |
| CogVideoX-5BOptimization Strategy=Vanilla Generation, Base Model=CogVideoX-5B2026.03 | 0.8013 | 0.4299 | 0.5419 | 0.2381 | 0.3857 | 41.98 | — | — | — | — | |
| DreamWorld2026.02 | 0.8011 | 0.5089 | 0.6182 | 0.1695 | 0.5507 | 52.97 | — | — | — | — | |
| GPT-4oOptimization Strategy=Best-of-N & VLM-Rating, Number of candidates (N)=5, Base Model=CogVideoX-5B2026.03 | 0.8005 | 0.4819 | 0.5706 | 0.2825 | 0.4103 | 46.15 | — | — | — | — | |
| GPT-4o (VQQA)Optimization Strategy=VQQA, Optimization rounds=4, Base Model=CogVideoX-5B2026.03 | 0.7995 | 0.5547 | 0.5792 | 0.3041 | 0.4692 | 48.18 | — | — | — | — | |
| VQAScoreOptimization Strategy=Best-of-N, Number of candidates (N)=5, Base Model=CogVideoX-5B2026.03 | 0.7985 | 0.5151 | 0.5792 | 0.3096 | 0.4249 | 46.95 | — | — | — | — | |
| Gemini-3-ProOptimization Strategy=Best-of-N & VLM-Rating, Number of candidates (N)=5, Base Model=CogVideoX-5B2026.03 | 0.7969 | 0.4796 | 0.5764 | 0.2879 | 0.3105 | 44.76 | — | — | — | — | |
| Wan 2.2-A14BScale=A14B2026.02 | 0.793 | 0.516 | 0.595 | 0.454 | 0.69 | — | — | — | — | — | |
| GPT-5.1Optimization Strategy=Best-of-N & VLM-Rating, Number of candidates (N)=5, Base Model=CogVideoX-5B2026.03 | 0.7883 | 0.4834 | 0.5649 | 0.2981 | 0.3388 | 45.28 | — | — | — | — | |
| VideoJAM2026.02 | 0.7868 | 0.4932 | 0.6489 | 0.1633 | 0.5242 | 52.33 | — | — | — | — | |
| VideoScore2Optimization Strategy=Best-of-N, Number of candidates (N)=5, Base Model=CogVideoX-5B2026.03 | 0.7809 | 0.4592 | 0.5505 | 0.2692 | 0.3546 | 43.97 | — | — | — | — | |
| Wan2.1-T2V-1.3BFine-tuned=true2026.02 | 0.7709 | 0.4313 | 0.628 | 0.1841 | 0.5451 | 51.18 | — | — | — | — | |
| Wan2.1-T2V-1.3B2026.02 | 0.7609 | 0.4592 | 0.5917 | 0.1681 | 0.5585 | 50.77 | — | — | — | — | |
| Summer-22BScale=22B2026.02 | 0.745 | 0.387 | 0.622 | 0.311 | 0.629 | — | — | — | — | — |